Retention data can tell you how many students in a defined starting group are still studying at a given date, and completion data how many have finished within a given time. On their own, they cannot tell you why students left, whether leaving was a failure, or whether the institution caused the outcome. A student who transfers, pauses for family reasons or takes a skilled job may count as a "loss" in one measure and a success in another. Useful retention evidence states its cohort, time window and treatment of transfers and part-time students, and follows departing students to where they went.
Six terms that are often confused
Much of the confusion in retention reporting comes from using one word for several different things. The definitions below are drawn from the official sources cited in this article; where agencies differ, the difference is the point.
| Term | What it usually counts | Example of an official definition | What it misses |
|---|---|---|---|
| Retention | Students still at the institution where they started | US National Center for Education Statistics (NCES): first-time undergraduates who return to the same institution the following fall | Students who continue elsewhere |
| Persistence | Students still enrolled anywhere | National Student Clearinghouse Research Center: students who remain enrolled at any institution | Whether they are still in the same program or field |
| Continuation | Students still studying or already qualified after a fixed period | Office for Students (England): one year and 15 days after starting, or two years and 15 days for part-time students | What happens after the observation date |
| Completion | Students who have gained a qualification within a time window | Office for Students: qualified (or still continuing) four years and 15 days after starting, six years and 15 days for part-time students; OECD: graduation at any tertiary level within the theoretical duration, plus one year and plus three years | Late completers outside the window; completions at another institution in some measures |
| Transfer | Students who leave one institution or program and enroll in another | NCES graduation rates count students who transfer without completing as noncompleters, even if they finish elsewhere | Often invisible to the sending institution |
| Interruption or stop-out | A pause followed by possible return | No common definition in the official sources reviewed here; a student counted as withdrawn may later return | Returns that occur after the reporting window closes |
Withdrawal and dropout sit beside these terms. In this article, withdrawal means the administrative event of leaving a course or institution, and dropout means leaving higher education without a qualification, the sense used in most of the research reviewed below. A withdrawal is not necessarily a dropout.
One cohort, several destinations
Official data that report all outcomes for one cohort show how much a single rate hides. NCES reports that among first-time, full-time students who started at US two-year institutions in fall 2017, about 34 per cent completed a certificate or associate degree within 150 per cent of normal time, 14 per cent had transferred, 10 per cent were still enrolled at their first institution and 42 per cent had left without completing or transferring. A completion rate of 34 per cent and a "did not complete" figure of 66 per cent are both accurate, but the second combines transfer, continued study and departure.
The gap between institution-level and system-level measures is visible in the National Student Clearinghouse Research Center's report on students who first entered college in fall 2024 (released 25 June 2026). Nationally, 77.1 per cent were still enrolled somewhere in fall 2025 (persistence), but 69.1 per cent were still at their starting institution (retention). The 8.0-point difference consists of students who were enrolled in their second fall, but not where they began. From the starting institution's records alone, they look like losses.
The OECD shows the same pattern across countries. In Education at a Glance 2025, on average across OECD countries and other economies, 43 per cent of full-time bachelor's entrants completed a degree at any tertiary level within the theoretical duration of their program, 59 per cent within one additional year and 70 per cent within three additional years. By the end of the theoretical duration, 42 per cent had graduated from a bachelor's program, 1 per cent had completed a short-cycle program, 38 per cent were still enrolled and 20 per cent had left tertiary education (the published figures sum to 101 because of rounding). The headline "on-time" rate therefore understates eventual completion substantially.
Why rates from different systems do not line up
Even before any question of quality, retention and completion figures differ because of how they are built:
- Cohort definition. NCES graduation rates use first-time, full-time entrants and adjusted cohorts that exclude, for example, students who left for military service. The OECD chapter covers only students who entered full time, while noting that about 30 per cent of tertiary students across OECD countries were enrolled part time in 2023.
- Time window. One year and 15 days, the following fall, 150 per cent of normal time, or theoretical duration plus three years all produce different numbers from the same students.
- Treatment of transfers. Some measures count completion anywhere; others count only completion at the starting institution.
- Calculation method. The OECD describes true cohort rates, which follow an entry cohort, as the preferred method, and states that the cross-cohort estimates used in earlier editions were not comparable and often overestimated completion.
- Population. An institution serving older, part-time or working students will show different rates from one enrolling recent school leavers full time, whatever the quality of either.
A retention percentage quoted without its cohort, window, transfer rule and population cannot be compared with another one.
What research shows about why students leave
Many causes, at several levels
Behr, Giese, Teguim Kamdjou and Theune (2020, Review of Education) review research on dropping out of university across disciplines. According to the abstract, they describe withdrawal as the result of a long decision-making process and a complex interaction of determinants, which they group into three levels: the national education system (for example, financing policy), the institution (for example, type of institution or teaching quality), and the students themselves, both before study (such as type of secondary school) and during it (such as working while studying). The grouping matters for interpretation: only some of these determinants are under an institution's control.
What departing students say
Greenland and Moore (2022, British Journal of Educational Technology) interviewed 226 students who had dropped out of open online education provided through Open Universities Australia (127 women and 99 men, mean age 33.45). Employment was the most common main reason, given by 81 of 226 participants (35.8 per cent), including overestimating their capacity to balance work and study. Asked what the university could have done to prevent their dropout, 190 of the 226 (84 per cent) first answered "nothing", describing the decision as their own and beyond the institution's control. With further probing, however, students identified 19 kinds of intervention in five broad areas that might have helped, many addressing personal and work circumstances that institutions often treat as unavoidable. The authors note that the study covers one non-traditional online cohort in Australia, conducted before the COVID-19 pandemic.
The finding cuts both ways. Many departures are driven by life outside the university, so a raw dropout rate overstates institutional failure. Yet students' first answer ("nothing could have helped") also understates what institutions can do, which is why exit evidence needs follow-up questions rather than a single tick box.
When leaving is not failure
The OECD states that "non-completion or delayed completion does not always reflect failure", listing switches of field, pauses for personal circumstances and early job opportunities among the reasons students leave.
Research that follows non-completers supports this. Tieben (2024, Empirical Research in Vocational Education and Training) uses retrospective life-course data from the German National Educational Panel Study (starting cohort 6). Of an analytic sample of 3,567 people who entered higher education aged 17 to 30, 922 (about a quarter) did not complete their initially chosen program. Within five years of non-completion, 55.3 per cent of non-completers with highly educated parents started another higher education program, compared with 40.3 per cent of those from less educated backgrounds. Among non-completers who already held a vocational qualification, 52.0 per cent moved into skilled employment, compared with 11.6 per cent of those without one. An earlier study by Tieben (2020, Research in Higher Education), read from its abstract, separates non-completion of the first program, transfer and dropout from higher education, and reports that after non-completion, traditional students were more likely to transfer while students with vocational qualifications more often left higher education.
The implication is that the same "departure" can be a transfer, a delayed completion, a move into skilled work or a genuine loss. These destinations also differ by social background, so a single attrition rate can hide inequalities in who gets a second chance.
Part-time, distance and working students
Students who study part time, at a distance or while working full time face different risks and are often measured differently. The Clearinghouse report notes that part-time starters in fall 2024 reached a second-fall persistence rate of 54.1 per cent, described as a decade high, compared with 77.1 per cent for the whole cohort.
Klinke, Kulle, Schreyögg and colleagues (2024, European Journal of Psychology of Education) analyzed institutional records for 5,796 students who started between 2006 and 2017 at a private German distance university of applied sciences. Part-time enrollment, older age, study interruptions and overdue payments predicted dropout, while a good match between the student's occupation and the program, and employer reimbursement of fees, predicted completion. Socio-demographic factors such as educational background, gender and nationality had little influence. The authors stress that the effects were small and that no causal conclusions can be drawn.
For evaluators, this means that comparisons between a distance or part-time provider and a campus-based full-time provider are comparisons between different populations and different time scales. Separate reporting by mode of study is the minimum.
What causal evidence shows about financial barriers
Most retention research is correlational, but financial aid is an exception with a body of causal studies. Nguyen, Kramer and Evans (2019, Review of Educational Research) conducted a systematic review and meta-analysis of the best available causal evidence. According to the abstract, 43 studies yielding 75 effect sizes estimate that grant aid increases the probability of persistence and degree completion by 2 to 3 percentage points, and that an additional $1,000 of grant aid improves persistence and attainment by 1.5 to 2 percentage points. The effect is real but modest: money matters for continuity, and it is not the whole explanation.
The Klinke study's finding on overdue payments points in the same direction, but it is a predictive association in one institution's data, not evidence that payment problems cause dropout.
The limits of early-warning prediction
Many universities now use institutional data to flag students at risk. The evidence suggests caution about how well such models work. In the Klinke study, a model using only information available before courses started classified 61.7 per cent of cases correctly, and a model adding information from after the start of study reached 70.6 per cent; the study identified poor grades and late submission of the first assignment as predictors of dropout. The authors note that prediction models perform better on the data they were built from than on future students, because they are prone to overfitting, and that the most important causes of dropout in distance learning, personal problems, are not visible in institutional records. Behr and colleagues also point to data-mining methods as a route to early-warning systems, but a flag is a prompt for contact, not a diagnosis.
A departure classification framework
The table below is an editorial framework proposed by this article. It is not the UNIRANKS methodology and does not describe how UNIRANKS evaluates institutions. It replaces a single attrition figure with destinations, each with the evidence needed to assign it and its likely interpretation.
| Destination | Evidence needed | Likely interpretation | Institutional question |
|---|---|---|---|
| Completed elsewhere or transferred and still studying | National enrollment or completion records, linked data | Continuity in the system; possibly a mismatch at entry | Was admissions advice or program information accurate? |
| Interrupted, with intention or right to return | Leave-of-absence records; re-enrollment tracked over several years | Temporary; outcome not yet known | Are return routes, credit retention and re-entry support in place? |
| Left for employment or vocational training | Exit interviews, follow-up surveys, administrative links where lawful | Not necessarily a failure; may reflect good outside options | Could flexible or part-time study have kept the option open? |
| Left for financial reasons | Exit data, hardship applications, payment records | Potentially addressable; causal evidence on grants exists | Was financial help known, timely and sufficient? |
| Left for health, family or caring reasons | Exit interviews with follow-up questions | Often outside institutional control, but not always | Were flexibility, deferral and support offered before departure? |
| Left after academic difficulty | Progression records, early assessment results | Most closely linked to teaching, support and admissions | Were early signals acted on, and how quickly? |
| Destination unknown | None | Cannot be interpreted | How can the unknown share be reduced? |
The share of departures in the last row is itself a quality indicator for the institution's data: the larger it is, the less any attrition figure means.
Our interpretation
Three conclusions follow. First, retention and completion figures are measures of continuity under specific rules, not measures of institutional quality. The NCES, Clearinghouse and OECD data each show large differences between "left this institution" and "left higher education", and between "on time" and "eventually". Second, the research points to departure as a process with several layers of causes, many outside the institution, and to destinations that differ by social background. Treating every departure as a failure misreads students who transfer, pause or take a skilled job; treating every departure as unavoidable misreads students who might have stayed with timely financial help, flexibility or contact. Third, the most informative institutional evidence is not a lower attrition rate but a better-explained one: departures classified by destination and reason, reported separately for part-time, distance and mature students, with the unknown share stated.
Students' perceptions of their experience are a related but separate question, discussed in How Should Universities Measure Student Experience? and, for participation and involvement, in Measuring Student Engagement Beyond Attendance.
Practical steps
For universities
- Publish the rule with the rate. State the cohort, the observation date, the treatment of transfers and part-time students, and whether the figure is a true cohort measure.
- Report persistence alongside retention where national data allow, so that students continuing elsewhere are visible.
- Classify departures by destination, using the categories in the framework above, and report the unknown share.
- Ask exit questions twice. Follow the first answer with specific prompts about work, money, flexibility and support, since students' first response may understate what could have helped.
- Keep the door open. Track interrupted students for several years, retain credit and make re-entry simple; a return after the reporting window is still a return.
- Use risk flags for outreach only. Treat predictive scores as prompts for human contact, check their accuracy on new cohorts and never use them to restrict access.
- Check financial support timing. Monitor payment problems and hardship applications as early signals, given the causal evidence that grant aid improves persistence.
For evaluators and researchers
- Compare only rates with the same cohort definition, window and transfer rule, and report mode of study separately.
- Read low retention at institutions serving part-time, older or distance students against their populations, not against full-time campus norms.
- Prefer destination-based follow-up over single attrition rates, and treat predictive associations in institutional data as hypotheses, not causes.
Limitations
This article draws on six peer-reviewed sources. Three were read in full (Greenland and Moore; Tieben 2024; Klinke and colleagues) and three from their abstracts only (Behr and colleagues; Tieben 2020; Nguyen, Kramer and Evans), so claims from those three are limited to what their abstracts state. The full-text studies each concern one country and, in two cases, one provider or one type of provider, so their percentages should not be generalized. The causal evidence on grant aid comes mainly from programs reviewed in one meta-analysis and is not specific to any institution. The official statistics come from the United States, England and the OECD, use different definitions and time windows, and are not directly comparable with each other. Several relevant studies were excluded because their publisher pages could not be opened without a bot check. The departure classification is an editorial framework and has not been validated as a measurement instrument.
Related reading on UNIRANKS
Research on satisfaction surveys, engagement, complaint resolution and belonging is collected in the Student Experience & Satisfaction research hub.
Editorial disclosure
UNIRANKS publishes university rankings, and its methodology includes a Student Experience & Satisfaction dimension. This article reviews independent research on student retention, continuity and completion data. It does not describe how UNIRANKS scores institutions.
Research References
- Behr, A., Giese, M., Teguim Kamdjou, H. D., et al. (2020). Dropping out of university: a literature review. Review of Education, 8(2), 614-652. https://doi.org/10.1002/rev3.3202 (verified 22 November 2026)
- Greenland, S. J., & Moore, C. (2022). Large qualitative sample and thematic analysis to redefine student dropout and retention strategy in open online education. British Journal of Educational Technology, 53(3), 647-667. https://doi.org/10.1111/bjet.13173 (verified 22 November 2026)
- Klinke, C., Kulle, K., Schreyögg, B., et al. (2024). Equal opportunities for non-traditional students? Dropout at a private German distance university of applied sciences. European Journal of Psychology of Education, 39(4), 4003-4024. https://doi.org/10.1007/s10212-024-00829-2 (verified 22 November 2026)
- Nguyen, T. D., Kramer, J. W., & Evans, B. J. (2019). The Effects of Grant Aid on Student Persistence and Degree Attainment: A Systematic Review and Meta-Analysis of the Causal Evidence. Review of Educational Research, 89(6), 831-874. https://doi.org/10.3102/0034654319877156 (verified 22 November 2026)
- Tieben, N. (2020). Non-completion, Transfer, and Dropout of Traditional and Non-traditional Students in Germany. Research in Higher Education, 61(1), 117-141. https://doi.org/10.1007/s11162-019-09553-z (verified 22 November 2026)
- Tieben, N. (2024). Destinations after higher education non-completion: the role of social background and pre-tertiary vocational qualifications. Empirical Research in Vocational Education and Training, 16(1), 9. https://doi.org/10.1186/s40461-024-00161-5 (verified 22 November 2026)
Sources
- Undergraduate Retention and Graduation Rates (Condition of Education, last updated May 2022). National Center for Education Statistics, U.S. Department of Education. https://nces.ed.gov/programs/coe/indicator/ctr (accessed 22 November 2026)
- Persistence and Retention: Fall 2024 Beginning Postsecondary Student Cohort (released 25 June 2026). National Student Clearinghouse Research Center. https://nscresearchcenter.org/persistence-retention/ (accessed 22 November 2026)
- Student outcomes data. Office for Students. https://www.officeforstudents.org.uk/data-and-analysis/student-outcomes-data/ (accessed 22 November 2026)
- Chapter B5. Who is expected to complete tertiary education? In Education at a Glance 2025: OECD Indicators. OECD. https://www.oecd.org/en/publications/education-at-a-glance-2025_1c0d9c79-en/full-report/who-is-expected-to-complete-tertiary-education_a1099e2e.html (accessed 22 November 2026)
